Reviewer: Chakchai So-in
Date: 9-13-2007
How would you rate this paper, relative to others we have read? top 50%, but not top 25%
How would you rate your knowledge of the topic of this paper? familiar, but not expert
What problem or issue does the paper address? Why is it important?
The problem here is to design the networking protocol for future Internet regardless of network congestion and misbehaving flows. The TCP, an end-to-end congestion control, is one the most important networking protocol. What the researchers have tried to achieve is to either modify or build new protocol to compete with TCP for stability, performance, and fairness purposes. Generally, the TCP-friendly protocol problems are that lacking the protection from the misbehaving flows, flow-unfairness, and high goodput.
What are the main contributions of the paper and why are they important?
With Erasure Correction Codes (obtaining the goodput) and approximate fair dropping (fairness issue at the router) schemes, the authors proposed the protocol which can maintain all those issues above but not considering an end-to-end congestion control.
How significant are these contributions relative to previous work?
In general, I think it is a good idea to think out of TCP; how about there is no need to consider the network congestion but there are a lot of issues needed to be concerned.
Give detailed comments justifying your view of the paper.
Apart from coding overhead, computation and complexity, source coding delay, deassembler, tons of dead packets, and packet reordering, the main problem here is deployability. It is difficult to think about how to merge this protocol to the exiting protocol, TCP; how the traditional flow behaves? Although the authors said, only the access routers are needed to modify, considering the hierarchical network, I believe most of the routers are needed to be changed to support AFD because the assumption only access network is bottleneck is not quite always true. One big concern is that since the sender keeps sending the full rate, it is quite difficult to think if the content server keeps feeding with 10Mbps to me on 56Kbps unless there is some mechanism to negotiate the maximum rate such as the received window as in TCP. Moreover, since the ACK mechanism here is very important that directly impacts on delay and coding method, the ACK needs to be loss-free.